FIXED : Keithley 199 ohms readings never settle

[DIPLover]

I just got a Keithley 199 for 40$ and although the body is a bit roughed up, the face which was covered in sticker residue (including the actual display area...  ?) cleaned up nicely with 99% alcohol and now looks new without even a scratch. 

I have started to compare it to my other meters to get a feel for it's calibration state and so far DCV and DCA are very good for the ranges I can test. (For example, after appropriate warm-up time etc. the 2.5V output of an AD584 reference reads 2.49945 vs my calibrated Keithley 196 which reads 2.499484, 10V out reads 10.0001 vs 10.00036 on the 196). Those are the ranges I use the most by far so I am quite satisfied.

DCA on the 3A scale is also very close to the 196 so no complaints. I got this for the superb display and don't intend to make any critical measurement with it.

But there is a problem with the ohms.

I picked a series of 1% resistors in 18, 180, 1.8K, 18K, 180K and 1.8Mohm values and measured them with several meters.
I compared the 199 to the average of all my other good meters, and to the 196 alone.
The 199 always reads high and is off by 8%, 0.3%, 3%, 4.5%, 5.5% and 6.8% on the different ranges.

That's a lot, but what is most alarming, is that the readings on the 199 never settle. Even on the 196, they will settle after several seconds down to the last digit. On the 199, the last 3 (sometimes 4) digits are never stable and will continue to change forever.
Except when I measure the 1.8M resistor using the 30Mohm range. Then the value is bang on (within 0.012% of the 196) and stable.

What could be the cause of this?

****** EDIT on 2019-03-20 ******
In the end my choice of 18 series resistor for the original test was very bad, since the calibration point for this unit are at 190, 1.9K, 19K etc. Thus being very close the errors didn't look too bad and I was more concerned with the jumpiness. It turned out all ranges were WAY out of wack except 30M and 300M.

In the end I replaced transistors Q16 BJT NPN with a 3904 (it was leaky). This fixed the 300, 3K and 30K ohms ranges.
I also ended up replacing Q11 N-Channel JFET with a pn4901 (failed short, all three pins reading the same 4.5+ Volts when it should have been closed and gate was receiving -14.8 V control signal). This fixed the 300K and 3M ranges.
I also had to recalibrate all ohms ranges except 30M and 300M since they had been calibrated (some by me) with the faults.
*************************************
It was a fun repair and I owe my thanks to Kleinstein. 

[Vgkid]

Have you tried using your K196 to measure the voltage across the resistor that is being measured by the 199. That could help point you if there is a current source problem. Likewise have you put a resistor across only the voltage terminals ou the 199 , in order to see if there is a leakage problem(doesn't seem like it is , but doesn't cause any harm).

[DIPLover]

Well, after a very busy year, finally got some time to investigate this properly...

It turns out the Keithley 196 uses the same circuit as the 199 for measuring ohms, with the same reference voltage and resistor values, so we can use it to get expected values. (see attached simplified diagram).

According to the manual, measured resistance should be equal to :

Rref * Vsense
-----------------
       Vref

Where Vref is 5 V and Rref depends on selected range :
300Mohm : 10 M
 30 Mohm : 10 M
  3 Mohm :   1 M
300 Kohm : 100 K
 30 Kohm :   30 K
   3 Kohm :     2 K
300 ohm :     2K

I setup a new test with a 5K precision resistor (4.9984K according to vendor).
I measured the voltage across the resistor for both the 196 and the 199 for the different ohms ranges of both DMMs.

Range ohms	Ohms 199	Voltage 199	Ohms 196	Voltage 196
300 M	000.004 M	0.0025 V	000.005 M	0.0025 V
30 M	00.0049 M	0.0025 V	00.00508 M	0.0025 V
3 M	1.19xxx M	0.2525 V	0.005006 M	0.0252 V
300 K	123.xxx K	0.2942 V	005.0000 K	0.2183 V
30 K	13.xxxx K	0.2641 V	04.99874 K	0.6965 V
3 K	2.75xxx K	1.1543 V	Overflow	3.019 V
300	overflow	1.1543 V	Overflow	3.020 V

I replaced digits with "xxx" in the table when they fluctuate and never stabilize.
We can see that the two highest ranges on the 199 give proper stable readings with correct values and correct voltage.

Voltage readings on 199 suggests effective VRef is measured as :
~99 K instead of 1 M on 3 Mohm range
~85 K instead of 100 K on 300 Kohm range
~95 K  instead of 30 K on 30 Kohm range

What kind of failure could cause behavior like that?
What should I investigate next?

[GregDunn]

I had a similar issue with an HP 34702A.  I would look at Q12; it seems to be the main current control for the ohms section and if it's leaky or damaged it could cause varying issues on multiple ranges.

[Kleinstein]

The problem could be due to a problem with the control signals (e.g. gate to Q3, Q11, Q12). There is also a chance that Q3, Q11 or U22 might be broken.

[DIPLover]

Thanks guys, I have started by checking Q12 and its control, U42B (which is an LM339 quad comparator).
Please see attached page of the actual schematic.
Q12 is located in zone C4.

U42 GND pin 12 is set to -10V
U42 Vcc pin 3 is set to +5V
U42 Input 1 pin 6 is set to +2.6V correctly
U42 input 2 pin 7 is controlled by the set Ohms function and range

Pin 1 of U42 (Q12 gate control) vs range
300Mohm : ON (-15V)
 30Mohm : ON
  3Mohm : ON
300Kohm : ON         
 30Kohm : OFF (0.5V)
  3Kohm : OFF       
300Ohm  : OFF         

Q12 drain to source voltage vs range
300Mohm : 0.0010V
 30Mohm :0.0010V
  3Mohm : 0.0010V
300Kohm : 0.0010V       
 30Kohm : 2.43V
  3Kohm : 1.98V       
300Ohm  : 1.98V

To me this seems to indicate Q12 and U42 are working fine together.
Or am I missing something obvious?
 
I am a digital engineer and have very little experience troubleshooting individual transistors so any pointers as to what should be tested and how to test it is greatly appreciated.

[Kleinstein]

Some of the readings are a little odd: Assume the gate voltage of Q12, when ON should be at +5 V and when OFF should be at -10 V.
The drain to source readings at least make sense and suggest this FET is working.

In the 3 M range seems to provide too much current to the resistor. AS the 30 M range kind of works Q13 and Q3 seem to be able to isolate.
Chances are that the extra current comes from Q13. The voltage-follower U20A could be a source a problem that would effect several ranges: too high a voltage here would cause gate current to flow.

[DIPLover]

Indeed I transcribed my notes wrongly.

I was measuring pin 1 of LM399 relative to pin 12 (-10V)

Correct Table is :
300Mohm : ON (+15V) (i.e. +5 Volts from analog ground)
 30Mohm : ON
  3Mohm : ON
300Kohm : ON         
 30Kohm : OFF (+0.5V) (i.e. -9.5 Volts from analog ground)
  3Kohm : OFF       
300Ohm  : OFF   

The actual gate voltage, measured at the transistor side of R52 is
300Mohm : 4.625V
 30Mohm : 4.625V
  3Mohm : 4.625V
300Kohm : 4.625V         
 30Kohm : -8.754V
  3Kohm : -8.754V       
300Ohm  : -8.754V 

[DIPLover]

From the simplified schematic and list of Rref for the different ranges, I built a truth table for the transistors :

Range           Q12     Q13     Q3      Q11
300 Mohm        1       0       0       0
 30 Mohm        1       0       0       0
  3 Mohm        1       1       0       0       
300 kohm        1       1       1       0
 30 kohm        0       0       0       1
  3 kohm        0       0       0       1
300  ohm        0       0       0       1

Q12 measurements agree with the TT. Moving on to Q13 as suggested by Kleinstein
Q13 is controlled by U2A, another LM339. This one uses -15V for pin 3 (gnd)

Measured U2 supply voltages :
Pin 12 (GND) : -15.058V
Pin  3 (Vcc) : 5.005 V

Pin 4 is a correct +2.6 V reference voltage (measures 2.59078V)

Pin 2 of U2 (Q13 gate control) vs range
300Mohm : -15 from analog ground)
 30Mohm : -15
  3Mohm : 0.310*
300Kohm : -15*   
 30Kohm : -15
  3Kohm : -15
300Ohm  : -15

Interestingly, Q3 is controlled by another comparator in the same LM339 chip (U2B).
Pin 1 of U2 (Q3 gate control) vs range
300Mohm : -15 from analog ground)
 30Mohm : -15
  3Mohm : -15
300Kohm : 0.362*
 30Kohm : -15
  3Kohm : -15
300Ohm  : -15

So although the supplies of LM339 designated U2 are correct, it seems incapable of giving a
strong positive output. (Looking at the datasheet, I see It outputs logic low or High-Z and needs
pullups for logic high. R93 was an obvious pullup for U42, nothing like that for U2)

R20D, R20B and R20A pull the outputs of U2 to pin1 of U20A, an LF442 dual opamp.

Pin 1 of U20A vs range
300Mohm : 0.391 V
 30Mohm : 0.391
  3Mohm : 0.313
300Kohm : 0.365
 30Kohm : 0.328
  3Kohm : 1.436
300Ohm  : 1.436

So U2 seems to operate correctly given what pullups it's got.

V+ of U20A (pin 8 ) : +14.991 V
V- of U20A (pin 4) : -15.057 V
Positive input (pin 3) :
300Mohm : 0.391 V
 30Mohm : 0.391
  3Mohm : 0.313
300Kohm : 0.365
 30Kohm : 0.328
  3Kohm : 1.436
300Ohm  : 1.436

Voltage follower U20A appears to do it's job. But why are the inputs so low?
There is a lot of stuff going on that node, not sure I can trace it

[DIPLover]

I notice Q5 and Q6 in zone D5 are on the same line.
I do not know this configuration.
Q6 has it's gate permanently wired to -4.3V and Q5 has both source and drain permanently to 5V.
Since it's the closest I could find to a +5V on that line, I measured :

Q5 gate voltages for ranges :
300Mohm : 0.00218 V
 30Mohm : 0.00218
  3Mohm : 0.248
300Kohm : 0.289
 30Kohm : 0.259
  3Kohm : 1.134
300Ohm  : 1.134

These are very close to the voltages measured across the tested resistor in post #3.

[wn1fju]

I had some strange ohms readings on my unit and it turned out to be Q16 - a protection transistor hung off the ohms line.  Lift it out (it isn't necessary anyway)
and see if anything changes.

[Kleinstein]

I would not expect Q16 to be the problem, as the high ohms ranges seem to work, and there is too much test current in one range.

My suspicion would be with Q11.

The  case in the 3 M range might give a clue: having only 0.25 V at the DUT (5 K) and reading 1.2 M would suggest the reference reading to the ADC is only at some 0.2 V, so much high and with quite a lot of voltage lost, as the sum is much below 5 V.

Most of the swithes are also used in other ranges, e.g. as a voltage divider for the higher voltage ranges. So if these work correct, Q3 and Q13 should work Ok.

Q5 and Q6 are used as a kind of low leakage diode. The readings suggest that not much current is flowing through R13 - which is a good thing. So Q5 and Q6 are not causing trouble.

The LM339 has open collector outputs and can thus only draw the signal down. To turn on a JFET it needs zero source-drain voltage. This level is usually supplied through a high (e.g. 100K, the 1 M at the gate are more for protection during transients and to slow down switching) resistor. So when on the gate voltage depends on the signal level present.

[DIPLover]

I had measured Pin 13 of U2, which is the Q11 gate control. For some reason I didn't post them... 

Pin 13 of U2 (Q11 gate control) vs range
300Mohm : -15 from analog ground)
 30Mohm : -15
  3Mohm : -15
300Kohm : -15
 30Kohm : 0.325 V*
  3Kohm : 1.425 V*
300Ohm  : 1.425 V*

Again the voltages are too low when the transistor should be ON.
The pullup on this line is R20A that goes to the same R20A LF442 voltage follower output, like the previous R20B, C and D pullups.

[Kleinstein]

Is the 5 V reference still OK at 5 V when a low Ohms range is selected. Too low a reference could explain some of the odd behavior.

The control voltage to Q11 looks reasonable. There is just too little test current flowing in the low ohms cases.

So maybe instead of testing a single point at different settings it might be better to set the meter to a non working range, like 30 K ohms and than measure the voltage at the points along the current path: the 5 V (VR), PIN10 and 11 of U22, drain of Q11 and the external 5 K resistor. 

[DIPLover]

Thank you for staying with me on this. I'll do as you propose and be back ASAP.

[DIPLover]

... and back ...

Testing whole path on 30K setting
5VR        : 5.09707
U22 pin 10 : 0.33376
U22 pin 11 : 0.34678
Q11 pin  1 : 0.33348
External 5K: 0.2641

[Kleinstein]

So it looks like most of the current drops across the 30 K resistor. For some reason some of the current flows a different path and not through the 5 K. The 2nd odd point is that the meter reads 13 K (from 1st. post). This suggests the voltage over the 30 K resistor does not find it's way to the ADC.   the way to the ADC is through U24. U24 C should be still On all the time. So the 0.347 V should also be at pin 11 of U24.

[DIPLover]

Indeed pin 11 of U24 reads 0.347 V when measuring the 5K resistor on the 30k range.


wn1fju :
Q16 emitter and base are at 0.00002 V
Q16 collector is at 0.32833 V

[DIPLover]

Pin 14 of U22D shows 5.09V
Pin 15 of U22D shows 0.334 V
No voltage drop across the 2K R26 resistor.

The working Keithley 196 using almost the same circuit (please see attached schematic, it is much better quality)
pushes 140 uA through the 5K resistor.
Our K199 only pushes 53 uA
 

[Kleinstein]

The voltage at Pin 11 of U24 suggests that there is not much current flowing this way and the signal is getting there.

The next step in the voltage reading path (pin 3 of U24) might need a scope to watch. It looks like somehow the signal is not getting through from pin 2 to 3.  This could be a defect in U24 or the control signal (pin1 of U24) is not right.  However this signal might be switching between levels. So the first step would be checking if PIN 1 is AC or fixed. If fixed the DMM would be enough to check pins 1 and 2.

For the current that is not coming to the 5 K resistor it might help to check with an infinite resistor (open). This may make it easier to see where the current goes, as the leakage would be the only path.  Q16 is a possible path (though not as the only problem). Still a brake-down voltage in the 0.3 V range would be odd for a bad transistor. There are not that many other paths that could carry away some 100 µA.
Q16 is used for protection as a low leakage diode, so for a test a different diode / transistor should be sufficient.

[coromonadalix]

hum dont want to barge in   but it is the schematics for a 196   not a 199

here's the 199

https://doc.xdevs.com/docs/Keithley/199/199_901_01D.pdf

EDIT :  oh shoot  they have the same measurement circuit   my bad  loll

[DIPLover]

Measured U24A.
Pin 2 is our familiar 0.347V DC signal.
Pin 1 is not DC but a slow train of wide 5V pulses.
Pin 3 passes the voltage in pin 2 when pin1 goes low.

@ coromonadalix :
It's OK they are not exactly the same, but K199 is clearly a reuse of the K196 circuit, with some simplifications. Also some of the switches are shuffled around a bit (for example K196's U24A is K199's U24C so the pin numbers are different) so the K196 schematic cannot be used alone. I attached the K199 schematic to post #5 in this thread.

[coromonadalix]

I may need to recalibrate my 196, unless i face the same problem as thoe OP            in the ohms, ac or dc volts  it never totally goes down to zero (input probes shorted)   few digits count,  last cal was done in 93 ?


I want to try understanding the data in my 196,  how the  pwm goes into the main logic system, but ive never done any of that, i have only a workbench logic sniffer in hand.

Data, clock , strobe, and the pwm  (measurements data)

Would be nice to hack an 1kv input in to the meter, dont need to be (hyper precise  say 0.5- 1% precision) to unlock an unused pin in the shifts registers,
or even try to control the measurement board section with an arduino 

[Kleinstein]

THe signals around U24A look good (given the 0.334 V to start with). So U24 is likely good.  So it is odd to get the wrong 13 K reading for a 5 K resistor. With to little current to the DUT one would normally expect to small a resistor to be read, more like 1.3 K.

For the path of the current, my guess now is on Q16  -  it would be a odd type of defect, but who knows. All the other switches seem to have high value resistors of 100 K or 1M in series - so no easy way to leak the missing 100 µA.
There might be a possible leak at the Ohms sense input, but there is also a 100 K resistor.

@coromonadalix:
The zero for the ohms ranges is usually a separate parameter to adjust, that is often possible by the user and without a full calibration. It can be a problem with the low ohms ranges in 2 wire mode. With 4 wire mode there should be essentially no zero error.

Adding a 1000 V or similar range has some danger to it. The other ranges of the meter don't look like they are well protected for a really high voltage. So some wrong handling could damage things. Also the switches may very well not up to 1000 V. If no high accuracy is needed, just get a cheap meter that does support the high voltage or build a separate meter (likely with a little less resolution) just for the 1000 V.

[DIPLover]

I removed Q16 (a 5818 NPN)

Voltage across 5K resistor with Q16 removed
300Mohm : 0.00253 V
 30Mohm : 0.00253
  3Mohm : 0.65407
300Kohm : 0.79073
 30Kohm : 0.70185
  3Kohm : 3.05337
300Ohm  : 3.05336

All ranges below 30 Mohm now read Overflow.

I then replaced it temporarily with a 3904 NPN
Voltage across 5K resistor with Q16 replaced
300Mohm : 0.00253 V
 30Mohm : 0.00253
  3Mohm : 0.65405
300Kohm : 0.79071
 30Kohm : 0.70184
  3Kohm : 3.05335
300Ohm  : 3.05335

Same behaviour.

The 3K and 300 ohm ranges now show similar voltages to the K196 across the 5K.
3M and 300M are much higher than K196.
30K is approximately the same, but reads overflow.